One of our latest projects at Velocity Research has been putting together our own low volume SMT manufacturing line to be able to quickly assemble small batches of circuit boards. The latest tool we have set up for this process is our reflow oven. After the solder paste is applied and the components are placed by the robotic pick and place machine, the boards will be placed in this oven to melt (aka reflow) the solder paste, electrically connecting the components to the board. This process requires precise control of the temperature to ensure that the solder melts evenly and no components are damaged. This is why a specialized reflow oven must be used.
The reflow ovens used at commercial board houses cost upwards of thousands of dollars and can quickly process large numbers of boards. On the other hand, we only need to be able to run small batches of boards so a smaller benchtop reflow oven would do the trick. After a lot of research, we decided on the T962A oven shown here:
This model is the right size for our application and is very popular in the maker community. However, our research did discover that there are a few known design flaws with the oven which can be quickly improved before first using it. The rest of this article will cover what we did to make these improvements and the results.
The first major issue many people have discovered the hard way is that parts of the oven’s insulation are secured using masking tape. Even though this tape is on the outside of the heat shielding, it still gets hot enough to burn and create a terrible melted glue smell even the first time the oven is brought up to temperature. The solution to this was to take apart the oven and replace all of the masking tape with Kapton tape which is designed to handle high temperatures. You can see the gold-colored Kapton tape in the image below:
The next fix was to replace the cooling fan for the main control board. The one that comes stock with the board is much louder than it needs to be and would make it a nuisance to run the oven around other people. However, it is a simple fix to wire in a similar-sized, quieter fan. We realized during the installation that the newer fan was designed to run on 5v instead of 12v like the original. So we did some testing on the bench to make sure that the fan could cool just as well and selected a resistor to wire in series to limit the current and allow the fan to be powered by 12v without burning out. Now The system is much quieter.
Cold Junction Reference:
This fix was important to the functionality of the oven. In order to maintain consistent and accurate temperature levels throughout the oven, there are a couple of thermocouples placed in the oven that the controller uses to measure the temperature.
The thermocouples take advantage of the Seebeck effect to measure temperature. For this to work there is a junction between two dissimilar conductors at the end of the thermocouple that is in the oven and another junction between these two conductors at the controller board, outside of the heated portion of the oven. A voltage differential is created that is proportional to the temperature difference between the two junctions. The problem with the way the oven is designed is that it assumes the cold junction is always at 20 degrees celsius, right below room temperature. However, this is not usually the case. The cold junction is still very close to the heated chamber of the oven and is around other power circuitry that causes it to heat up. Some users have found that there is an extra unused pin on the microcontroller that, with a firmware modification, can be used to interface with a simple one-wire digital temperature sensor. If this sensor is placed near the cold junction of the thermocouple, the controller can compensate for the temperature fluctuations and get a much more accurate measurement of the temperature in the oven. We used the Maxim Integrated DS18B20+T&R temperature sensor, which is shown on the reworked board below.
People have reported that the stock firmware is not as user friendly as it could be. Thankfully people have written updated firmware that is easier to use and implements the cold junction temperature reference. The motherboard has communication pins available that can be used to reprogram the microcontroller. After some trial and error, we were able to successfully reflash the new firmware and it worked perfectly! The link to the new firmware and an install walkthrough can be found here: https://github.com/UnifiedEngineering/T-962-improvements/wiki/Flashing-the-LPC21xx-controller
After making the modifications and reassembling the oven, the next step was to fire it up and calibrate the temperature control. The reading from the thermocouples is proportional to the temperature in the oven, but it may need to be scaled and shifted slightly. The updated firmware has settings to make these adjustments. To determine the necessary gain and offset value for each thermocouple, we measured the actual temperature in the oven using an accurate fluke multimeter thermocouple. We took many samples across the temperature range of the oven and plotted it vs what the controller thought the temperature was. Then we simply determine how much these values would need to be scaled and shifted so that their ratio was exactly 1:1, and then we saved those adjustments in the oven’s settings. This was done for each of the two thermocouples. After this, the oven could hold a temperature reading fairly accurately within a few degrees C range.
All that was left was to run a test board through the oven. We hand placed a few components on a blank circuit board and ran it through the oven. We made sure to use the correct reflow profile for the 63/37 Tin/Lead solder paste we were using, which is shown below:
After a couple of trials, adjusting the amount of solder paste used since we didn’t have a stencil, the oven worked perfectly! Now we know that once we begin assembling our own boards, we will be able to reliably reflow our solder joints.